Episodic Memory

Abstract

Episodic memory is the form of memory that preserves particular personal events, set in their place and time, and lets a person return to them in thought. Endel Tulving separated it from semantic memory and tied it to autonoetic consciousness, and later work has kept the distinction while finding the two systems more intertwined than once assumed. The hippocampus and medial temporal lobe are central to it, but whether memories depend on the hippocampus for as long as they last remains contested among standard consolidation, multiple trace, trace transformation and contextual binding accounts. Remembering is constructive and fallible, and the same simulation process appears to support imagining future events. Whether nonhuman animals possess episodic memory, and whether the construct should give way to a broader notion of event memory, are open questions.

Keywords: episodic memory, autonoetic consciousness, hippocampus, memory consolidation, constructive memory

Episodic memory is memory for events a person has personally lived through. In a review that looked back over three decades of work on the idea, Tulving presented it as a neurocognitive system in its own right, bound up with the self, with subjective time and with a particular kind of awareness he called autonoetic consciousness (Tulving, 2002). It is usually set beside semantic memory, the store of general knowledge, and a recent review judged the two to be intertwined in operation yet still distinguishable (Renoult et al., 2019).

This page sets out what episodic memory is and how it relates to semantic memory, the awareness that accompanies remembering, and the brain systems that support it. It then turns to the unresolved dispute over how episodic memories change with time, to the constructive and fallible character of remembering, to the link between memory and imagining the future, and to the evidence from animals, closing with a table of the rival accounts and their present standing.

What Episodic Memory Is

Episodic memory is defined less by what it stores than by how its contents are experienced. Tulving's mature account placed it among the memory systems as a separate neurocognitive system, closely tied to the self, to a sense of subjective time and to autonoetic consciousness (Tulving, 2002). That account was itself the product of revision. Surveying roughly 30 years of research, Tulving acknowledged that the concept had changed greatly since it was first proposed, and he devoted part of the review to criticisms of it (Tulving, 2002).

Within the wider taxonomy, episodic memory is a kind of declarative memory, one branch of long-term memory. Squire's synthesis of findings from rats, monkeys and humans separated declarative memory from nondeclarative forms such as skills and habits, priming and simple conditioning, and held that the declarative branch depends on the hippocampus (Squire, 1992). Henke proposed a processing-based alternative to this kind of taxonomy, in which memory systems are told apart by processing modes rather than by conscious awareness, which she judged a poor criterion; in her proposed model the hippocampus supports rapid binding whether or not the learner is aware of it (Henke, 2010). Episodic memory therefore sits under memory, its parent heading in MeSH, and beside semantic memory within the declarative branch, as Figure 1 shows.

The MeSH descriptor Memory, Episodic also files autobiographical memory and prospective memory under this heading.

Autobiographical memory has been studied with the Autobiographical Interview: across 5 life periods, older adults recalled fewer episodic and more semantic details than younger adults, a pattern that persisted after structured probing (Levine et al., 2002). Conway and Pleydell-Pearce treated such memories as transient constructions formed within a self-memory system (Conway & Pleydell-Pearce, 2000). Prospective memory showed no age deficit in two laboratory experiments with young and old adults, while age deficits appeared on the retrospective tests (Einstein & McDaniel, 1990).

Figure 1

Where Episodic Memory Sits Among Memory Systems

Episodic and Semantic Memory

The division between remembering events and knowing facts dates from 1972, when Tulving put forward the episodic-semantic distinction in his chapter on episodic and semantic memory in Organization of Memory (Tulving, 1972; Renoult et al., 2019), and he returned to the subject at book length in a later monograph on the elements of episodic memory (Tulving, 1983). Its standing today is best described as revised rather than overturned. Weighing the evidence nearly five decades on, Renoult and colleagues concluded that the two forms of memory are intertwined yet distinct: the line between them is less clear-cut than earlier work assumed, and the brain activity associated with each overlaps considerably (Renoult et al., 2019). Tulving had earlier noted how far the concept had moved from its original form and had engaged with criticisms of it (Tulving, 2002). A more radical revision, taken up in the Discussion, would set the episodic construct aside altogether in favour of event memory (Rubin & Umanath, 2015).

The clearest sign that the two systems can come apart comes from patients injured early in life. In a case series of three patients with early hippocampal injury, examined with magnetic resonance techniques, Vargha-Khadem and colleagues found global anterograde amnesia in which episodic memory was impaired while semantic memory was largely spared (Vargha-Khadem et al., 1997). The authors judged the two components only partly dissociable, since only episodic memory appeared to rely wholly on the hippocampus, while the patients' semantic attainment fell in the low average to average range (Vargha-Khadem et al., 1997).

Autonoetic Consciousness and Mental Time Travel

Tulving's distinctive step was to define memory systems partly by the awareness that accompanies them. In a 1985 paper he proposed three memory systems, procedural, semantic and episodic, alongside three kinds of consciousness, anoetic, noetic and autonoetic, and linked episodic memory with autonoetic consciousness (Tulving, 1985). In his later formulation, that awareness is the one that involves the self and subjective time (Tulving, 2002). By Tulving's account, the original evidence came from a case study of one amnesic patient and two simple demonstration experiments with normal subjects, testing recall and recognition of word-events, from which he offered speculations rather than firm conclusions (Tulving, 1985).

The idea was later carried into the brain. Wheeler, Stuss and Tulving proposed that the prefrontal cortex supports the autonoetic retrieval of episodes, framing remembering as a form of mental time travel, and presented the proposal as a preliminary theory (Wheeler et al., 1997).

Mental time travel then became the centre of a comparative argument. Suddendorf and Corballis treated the recall of past episodes and prospection of future ones as aspects of one capacity and asked whether that capacity is unique to humans; their review of comparative studies found no convincing evidence of mental time travel in nonhuman animals so far (Suddendorf & Corballis, 2007). Corballis later revised that earlier view in a brief commentary, arguing for evolutionary continuity instead and citing hippocampal activity in humans and rats as evidence (Corballis, 2013). The claim that mental time travel is uniquely human is contested, because work on food-caching birds, described below, has been read as evidence of episodic-like memory in another species (Clayton & Dickinson, 1998).

The Hippocampus and the Medial Temporal Lobe

Laboratory studies probe episodic memory through recall, in which a person reproduces studied material, and recognition, in which a person judges whether an item was studied. On the neural side, Scoville and Milner's 1957 paper concerned the loss of recent memory after bilateral lesions of the hippocampus (Scoville & Milner, 1957).

Squire later drew 3 separate literatures, on rats, monkeys and humans, into one synthesis. He concluded that the hippocampus is essential for declarative memory, but argued that its contribution is temporary, consisting in binding together sites in the neocortex (Squire, 1992). The second half of that claim is the subject of the consolidation debate in the next section.

Recognition need not rest on a single process. Dual-process accounts separate recollection, the retrieval of details of the original episode, from familiarity, a sense that an item has been met before without those details. Yonelinas reviewed 30 years of research on the nature of these two processes (Yonelinas, 2002). A later review by Eichenbaum, Yonelinas and Ranganath, drawing on work in humans, monkeys and rats, assigned recollection, but not familiarity, to the hippocampus, and considered the perirhinal and parahippocampal cortices among the other medial temporal contributors (Eichenbaum et al., 2007). The same authors noted that the medial temporal lobe's role in recognition remains debated, and they offered their suggestion that the parahippocampal cortex supplies spatial context only tentatively (Eichenbaum et al., 2007). The dual-process account of recognition is therefore best described as contested. Wixted and Squire made the rival case: in their view the hippocampus supports familiarity as well as recollection, and earlier contrasts between the two processes were confounded with memory strength, though they did not offer strength as the organising principle and noted that single-attribute tasks and face memory still needed testing (Wixted & Squire, 2011).

How Episodic Memories Change Over Time

How long an episodic memory relies on the hippocampus is an open theoretical dispute, and each of the main answers to it is contested. Four accounts are in play.

Standard, or systems, consolidation theory gives the hippocampus a time-limited job. On Squire's account the structure is essential for declarative memory, yet its role is temporary, binding neocortical sites together (Squire, 1992). A 2015 review by Squire and colleagues gave the theory its current form, in which the hippocampus guides a reorganisation of the neocortex that ends with the memory independent of the hippocampus, drawing on work on neural replay and sharp wave ripple activity; the review reported consolidation for episodic and semantic, spatial and nonspatial memory alike, while conceding that the findings on these points are inconsistent and theoretically disputed (Squire et al., 2015). The theory remains contested; systems consolidation is the explicit target of the contextual binding theory, which sets out to reconsider it (Yonelinas et al., 2019). Much of the argument concerns graded retrograde amnesia, the pattern in which loss after damage differs with the age of the memory; the contextual binding theorists judged such gradients rare after selective hippocampal damage and the few reported cases open to both accounts (Yonelinas et al., 2019).

Multiple trace theory took the opposite position. Reviewing lesion studies in humans and other animals, Nadel and Moscovitch argued that episodic and spatial memories need the hippocampal complex throughout their existence, and that how much is lost after hippocampal damage depends on the type of memory involved (Nadel & Moscovitch, 1997). Moscovitch and colleagues later set this within a hippocampus-centred component process model, first proposed in 1992, in which memories are transformed throughout life, a process they connected with schemas, gist and semantic knowledge and with differences along the hippocampal long axis; their review weighted recent work and gave limited coverage to emotion and the lifespan (Moscovitch et al., 2016). Sekeres, Winocur and Moscovitch then updated trace transformation theory around a gradient running along that long axis, from posterior to anterior hippocampus, together with the medial prefrontal and entorhinal cortices, and they treated reverse transformation as a possibility rather than an established finding (Sekeres et al., 2018). Even in its updated form (Sekeres et al., 2018), multiple trace and trace transformation theory is contested, because the 2015 restatement of standard consolidation described above still has the hippocampus guiding a reorganisation of the neocortex until memories no longer depend on it, though the authors of that restatement acknowledged unresolved empirical inconsistencies and theoretical disagreements (Squire et al., 2015).

Contextual binding theory offers a third answer: the hippocampus ties each item to the context in which it was met, and forgetting comes from interference between contexts, not from a hippocampal role that fades, although its authors grant that some findings fit both this account and systems consolidation (Yonelinas et al., 2019). Yonelinas, Ranganath, Ekstrom and Wiltgen proposed that the hippocampus binds each item to the context in which it was encountered, and that forgetting arises from contextual interference, with contextual drift also part of the picture (Yonelinas et al., 2019). The authors were candid about what their theory leaves unsettled. They granted that some findings suit both their account and systems consolidation, that the stronger effect of reactivation on weak memories is unexplained, that consolidation gradients in fear conditioning are mixed, that the hippocampus's part in memory outside the episodic kind is unresolved, and that neither account yet decides what cueing during sleep shows; their search of the neuroimaging literature was informal (Yonelinas et al., 2019). Contextual binding theory is accordingly contested as well.

No account has displaced the others. The standard view keeps its cross-species base, the transformation view keeps the lesion evidence that loss depends on memory type, and the contextual binding view keeps its account of forgetting, while some of the evidence fits more than one of them.

Constructive Memory and False Recall

How an episode is encoded and how it is later retrieved were tied together in Tulving and Thomson's 1973 paper on encoding specificity and retrieval processes in episodic memory (Tulving & Thomson, 1973). The principle's central idea, the match between the conditions of encoding and those of retrieval, was challenged by Nairne, who argued that this match relates to retention only correlationally, not causally: the link is usually positive, yet more match can help, make no difference or hurt, so match alone yields no unequivocal predictions about retention (Nairne, 2002). The principle is therefore best described as contested.

False memories can be produced reliably in the laboratory. Across two experiments using the Deese paradigm, Roediger and McDermott presented lists, of 12 words in the first experiment, whose members were associates of a related word that was never itself shown. On free recall tests and recognition tests alike, participants produced robust false recall and false recognition of those nonpresented associates (Roediger & McDermott, 1995).

This fallibility fits the view, developed in the next section, that remembering is itself a constructive simulation, the same process that builds imagined events (Addis, 2018).

Remembering the Past and Imagining the Future

Remembering and imagining appear to share their machinery. Schacter and Addis set out the constructive episodic simulation hypothesis in 2007, arguing that remembering the past and imagining the future overlap and draw on shared neural substrates, while cautioning that evidence on future events was still sparse and that the link between past and future was mostly unexplored (Schacter & Addis, 2007). In an event-related fMRI study of 16 healthy adults, 14 of whom were analysed, past and future events engaged a shared region of the left hippocampus, a result qualified by the small sample and by uncorrected thresholds (Addis et al., 2007). Addis later restated the hypothesis as the claim that remembered and imagined events are products of one constructive simulation process, which she related to the default mode network (Addis, 2018). A review by Schacter, Benoit and Szpunar found that a core brain network supports the simulation of future episodes, while noting that the contributions of its individual regions are still being separated (Schacter et al., 2017).

Patient evidence points the same way. Hassabis and colleagues compared patients with bilateral hippocampal amnesia with matched controls, asking both groups to imagine new experiences from short verbal cue scenarios. The patients' imagined experiences were impaired, notably in their spatial coherence, and the authors raised the idea that the hippocampus provides the spatial context for such scenes only as a possibility (Hassabis et al., 2007). Hassabis and Maguire offered a rival reading of the same overlap: on their account, what remembering and imagining share is better explained by scene construction than by self-projection, a view they argued for in an opinion article (Hassabis & Maguire, 2007).

Addis later set out a neurocognitive model of event simulation that revisits the notion of mental time travel (Addis, 2020). Together, this work treats memory of the past and thought about the future as two uses of one capacity, a view that also bears on the animal question below.

Episodic-Like Memory in Animals

Whether any nonhuman animal has episodic memory depends on what counts as evidence. Clayton and Dickinson had scrub jays cache food, including wax worms and peanuts, at visuospatially distinct sites and later recover it, in a design that turned on the perishability of the food. The jays' recovery behaviour showed that they remembered what they had cached, where and when, which the authors called episodic-like memory (Clayton & Dickinson, 1998). The qualifier carries weight: the criteria were behavioural only, so the result establishes memory for the content of an event in a non-human animal, not conscious recollection of it.

Suddendorf and Corballis took the sceptical side, concluding from comparative studies that no convincing evidence yet shows mental time travel in nonhuman animals (Suddendorf & Corballis, 2007). The question of whether mental time travel is uniquely human therefore remains contested. Table 1 summarises this and the other accounts discussed on this page.

Table 1

Accounts of Episodic Memory and Their Standing

Note. The Status column gives each account's present standing, and the citation in each row names the work on which that standing or claim rests.

Discussion

Some things have held. Episodic memory is still treated as a distinct capacity for remembering personal events (Tulving, 2002), and the hippocampus as essential to the declarative memory it belongs to, if only temporarily on the cross-species synthesis (Squire, 1992). The selective impairment of episodic memory after early hippocampal injury, partial though it was, points the same way (Vargha-Khadem et al., 1997). It is equally well established that remembering is constructive and fallible, since ordinary list learning yields robust false recall and recognition (Roediger & McDermott, 1995).

Other things have moved. The boundary with semantic memory, once drawn sharply, is now seen as less clear-cut, with overlapping neural correlates (Renoult et al., 2019). Moscovitch and colleagues, proponents of the transformation view, reported a growing tendency to treat episodic memory as transformed throughout life (Moscovitch et al., 2016). And memory has been tied to imagination, with remembering and imagining treated as one simulation process (Addis, 2018).

Three questions remain open, and they connect. The first is the hippocampus's role over time, where contextual binding and systems consolidation still compete and some findings fit both (Yonelinas et al., 2019). The second is the animal evidence, where behavioural criteria have been met but a 2007 review of comparative studies found no convincing evidence of mental time travel (Suddendorf & Corballis, 2007). The third is whether episodic memory is a natural kind at all. Rubin and Umanath proposed replacing the construct with event memory, a theory covering memory for laboratory, autobiographical and fictional events in which scene construction locates the self in space and time, a sense of reliving is not required, and a single memory can summarise several encodings (Rubin & Umanath, 2015). The proposal sharpens the other two questions: if reliving is not essential, behavioural evidence in animals might count as event memory without meeting Tulving's criteria for episodic memory. Settling these questions would require evidence that separates the rival predictions directly, such as following memories of different ages and types in the same patients, rather than findings that each account can absorb.

Current Directions

Recent work has concentrated on transformation, binding and simulation. On the consolidation front, the component process model treats episodic memory as transformed across the lifespan, with the hippocampal long axis given a central place (Moscovitch et al., 2016), and the updated trace transformation theory specifies a posterior-to-anterior gradient within the hippocampus working with prefrontal and entorhinal regions (Sekeres et al., 2018). Contextual binding theory challenges systems consolidation by recasting forgetting as contextual interference, while listing several points its authors regard as unresolved (Yonelinas et al., 2019).

The relationship between episodic and semantic memory is being redrawn as one of entanglement rather than clean separation (Renoult et al., 2019). On the simulation side, the constructive episodic simulation hypothesis has been restated as a claim that remembered and imagined events arise from the same simulation process (Addis, 2018), a neurocognitive model of event simulation has been put forward (Addis, 2020), and research on episodic future thinking is still working to separate what each region of the core network contributes (Schacter et al., 2017).

Commonly Confused With

Semantic Memory

Episodic memory holds events a person has personally experienced and is tied to autonoetic awareness of the self in subjective time (Tulving, 2002). Semantic memory holds general knowledge, and the two are intertwined in operation yet distinct, with boundaries less clear-cut than once assumed and overlapping neural correlates (Renoult et al., 2019). They can come apart: three patients with early hippocampal injury showed impaired episodic memory while semantic memory was largely spared, although the dissociation was only partial (Vargha-Khadem et al., 1997).

Common Misconceptions

Memories are stored like recordings and replayed exactly.

Studying lists of related words leads people to recall and recognise associated words that were never presented, and the effect is robust (Roediger & McDermott, 1995). Remembering has instead been described as a constructive simulation, the same process that produces imagined events (Addis, 2018).

The hippocampus is needed only for a short time after learning.

That is one side of an unresolved dispute. Standard consolidation theory does give the hippocampus a temporary role (Squire, 1992), but multiple trace theory holds that the hippocampal complex is required for the whole life of an episodic or spatial memory (Nadel & Moscovitch, 1997).

Only humans can remember events.

Scrub jays remember what they cached, where and when, although the evidence rests on behavioural criteria only and does not show conscious recollection (Clayton & Dickinson, 1998). Whether any nonhuman animal travels mentally in time is a separate question, on which a 2007 review of comparative studies found no convincing evidence (Suddendorf & Corballis, 2007).

Glossary

Autonoetic consciousness.

The self-aware form of consciousness that Tulving linked to episodic memory, involving the self and subjective time; his first statement of the idea was offered as speculation.

Constructive episodic simulation.

The hypothesis that remembered and imagined events are products of a single constructive simulation process; remembering past and imagining future events engage overlapping brain regions, whose separate contributions are still being worked out.

Contextual binding.

The contested proposal that the hippocampus binds items to the context in which they occurred, with forgetting attributed to contextual interference.

Declarative memory.

The branch of long-term memory, including episodic and semantic memory, that depends on the hippocampus and is distinguished from nondeclarative forms such as skills and habits, priming and simple conditioning.

Encoding specificity.

The principle, from Tulving and Thomson, that ties remembering to the match between encoding and retrieval conditions; contested on the ground that this match relates to retention only correlationally and can help, make no difference or hurt.

Episodic-like memory.

Memory for what happened, where and when, shown in animals such as scrub jays by behavioural criteria only, without evidence of conscious recollection.

Event memory.

A proposed replacement for the episodic memory construct, covering memories that need not involve a sense of reliving and that can summarise several encodings.

Familiarity.

In dual-process accounts of recognition, a sense that an item has been encountered before without retrieval of the episode; a process not attributed to the hippocampus.

Mental time travel.

Mentally re-entering past episodes or projecting future ones; whether nonhuman animals do so is contested.

Multiple trace theory.

The contested account holding that episodic and spatial memories rely on the hippocampal complex at every age, never outgrowing it.

Recollection.

In dual-process accounts of recognition, the retrieval of details of an earlier episode, a process attributed to the hippocampus; the wider medial temporal lobe's role remains debated.

Semantic memory.

Memory for general knowledge, distinguished from episodic memory since 1972 and now regarded as intertwined with it yet distinct.

Systems consolidation.

The contested view that the hippocampus plays a temporary role in declarative memory, binding neocortical sites together.

Trace transformation theory.

The contested view that memories are transformed throughout life, updated to include a gradient along the long axis of the hippocampus.

Key Researchers

Donna Rose Addis

Senior scientist at the Rotman Research Institute, Baycrest, and Canada 150 Research Chair at the University of Toronto, who argues that remembering and imagining are products of one simulation process. ORCID - Wikidata - Wikipedia - Faculty Page

Nicola S. Clayton

(b. 1962). Professor of Comparative Cognition at the University of Cambridge, who with Anthony Dickinson showed that scrub jays remember what they cached, where and when. Wikidata - Wikipedia - Faculty Page

Brenda Milner

(b. 1918). Neuropsychologist at the Montreal Neurological Institute, McGill University, who with William Scoville reported the loss of recent memory after bilateral hippocampal lesions. Wikidata - Wikipedia - Faculty Page

Morris Moscovitch

Max and Gianna Glassman Chair in Neuropsychology and Aging at the University of Toronto and senior scientist at the Rotman Research Institute, Baycrest; with Lynn Nadel he proposed that the hippocampus stays involved in episodic memory for as long as the memory lasts. Wikidata - Wikipedia - Faculty Page

Daniel L. Schacter

(b. 1952). William R. Kenan, Jr. Professor of Psychology at Harvard University, whose fMRI study with Addis and Wong compared remembering past events with imagining future ones (Addis et al., 2007), and whose work traces the core brain network that supports imagining future events. ORCID - Wikidata - Wikipedia - Faculty Page

Larry R. Squire

(b. 1941). Emeritus Distinguished Professor at the University of California, San Diego, whose synthesis of work with rats, monkeys and humans argued that the hippocampus is essential for declarative memory. Wikidata - Wikipedia - Faculty Page

Endel Tulving

(1927–2023). University Professor Emeritus at the University of Toronto who proposed the distinction between episodic and semantic memory and tied episodic memory to autonoetic consciousness. Wikidata - Wikipedia - Faculty Page

Andrew P. Yonelinas

Professor of Psychology at the University of California, Davis, who reviewed the evidence for recollection and familiarity as separate processes and coauthored the contextual binding theory of episodic memory. Faculty Page

Frequently Asked Questions

What is episodic memory?

Episodic memory is memory for events a person has personally experienced. Tulving described it as a distinct neurocognitive system bound up with the self, subjective time and autonoetic consciousness (Tulving, 2002).

How does episodic memory differ from semantic memory?

Semantic memory holds general knowledge, whereas episodic memory holds particular events. Current reviews treat the two as intertwined yet distinct, with boundaries less clear than once assumed and considerable overlap in their neural correlates (Renoult et al., 2019).

What is autonoetic consciousness?

It is the kind of awareness Tulving attached to episodic memory, one of three kinds of consciousness he proposed alongside anoetic and noetic consciousness. He offered the scheme as speculation, based on one amnesic patient and two simple demonstration experiments (Tulving, 1985).

What does the hippocampus do in episodic memory?

A synthesis of findings in rats, monkeys and humans concluded that the hippocampus is essential for declarative memory, the broader class to which episodic memory belongs (Squire, 1992).

Do memories stay dependent on the hippocampus?

This is disputed. One answer, multiple trace theory, holds that an episodic or spatial memory never stops depending on the hippocampal complex, however old the memory is, with the extent of loss after damage depending on the type of memory (Nadel & Moscovitch, 1997).

Why do people remember things that did not happen?

Memory is constructive. In the Deese paradigm, studying lists of 12 related words produced robust false recall and false recognition of associated words that were never presented (Roediger & McDermott, 1995).

Do animals have episodic memory?

Scrub jays remember what food they cached, where and when, which meets behavioural criteria for memory of events, although such criteria cannot show conscious recollection (Clayton & Dickinson, 1998).

How does remembering relate to imagining the future?

According to the constructive episodic simulation hypothesis, remembered and imagined events are products of one constructive simulation process (Addis, 2018).

References